河床演变野外采集源数据滤波方法研究

Research on filter methods for field data of riverbed evolution

随着现代测控技术的迅速发展 ,由野外实测而得到的数据已成为一种非常重要的和极具潜力的河床冲淤演变信息来源。为确保河床演变分析源数据的可靠性 ,结合源数据的特点 ,对其滤波方法进行了深入研究。本文在简述现代河床演变参数实测技术的基础上 ,考虑现代水下地形测量的动态特征 ,提出了适合于地形数据滤波的趋势面法和基于M估计的抗差滤波法 ,以及适合于动态水位观测数据的卡尔曼滤波模型 ,对模型优缺点、及关键参数的确定进行了深入分析 。

With the rapid development of advanced surveying and controlling technologies, data acquired from the field is the very important and the most potential source for riverben evolution. Generally, original data observed from the field is thought to be accurate. However, for riverbed evolution data, sometimes, outstanding gross error is merged in original data. In order to eliminate these error effect in data processing and spatial analysis of riverbed data, trend_area filter and robust filter are introduced in this paper. According to research, trend_surface filter is easy in application, and adapts to outstanding gross error checking, but it isn't sensitive for small gross error. Of course, choice of degree number of polynomial and filter threshold is very important for filter effect of trend_area filter. In order to overcome above limitation, a kind of robust filter based on M_estimate is given in the paper. The robust filter can be used for finding and marking both outstanding gross error and small gross error. In the research of robust filter, the determination of some important parameters, including RMS(Root Mean Square) and initial value are researched, and ideal initial value and model for determination of RMS are given, and are proved by experiment to be right. Integrating with its character, Kalman filter is used for eliminating water_lever data error. The satisfying result is acquired as well.